Fugu Max

Sakana AI released Fugu Max alongside Fugu Ultra v2, with the GIGAZINE English coverage explicitly describing Fugu Max as the cost-effective entry in the Fugu multi-agent orchestration line. According to the article, Fugu Max uses the same task-routing approach as the rest of the family, dispatching sub-tasks across a The article lists Fugu Max API pricing at roughly $2 per million input tokens, $0.25 per million cached input tokens, and $6 per million output tokens on a pay-as-you-go or subscription basis, positioning it below Fugu Ultra v2 ($5 / $0.50 / $30 per million tokens respectively) and reinforcing Sakana's claim that Fugu

Eesel AI's explainer describes Fugu Max as a model that calls other models, positioning it as the fourth model in the Sakana Fugu family behind a one-line API parameter switch on the same OpenAI-compatible endpoint. The family lineup covers Fugu for everyday balanced performance, Fugu Ultra (now v2) for peak answer qua The piece explains that when you invoke Fugu Max, you are not talking to one network of weights but a coordinator that decides per request who should do the work, expanding to a large pool of open-weight and specialized models and sending each sub-task to the cheapest model that can actually handle it. Sakana is quoted

TechJack Solutions' AI Brief confirms the September 11, 2026 launch of Fugu Max alongside Fugu Ultra v2 and corroborates Sakana's pricing from a second source: Fugu Max at $2 per million input tokens and $6 per million output tokens with a 1M-token context window, as confirmed by both Sakana's announcement and OpenRout The brief notes vendor claims that Fugu Max's output pricing is 40–60% lower than Sonnet 5, GPT 5.6 Terra, and Kimi K3 (flagged as unverified vendor claims), and adds internal benchmark figures of 48.3 on Chartography and 74.3 on DeepSWE for Fugu Ultra v2. Both models share the multi-agent orchestration architecture an

DataNorth's independent write-up of the September 11, 2026 launch consolidates the Fugu Max technical specification table: $2 per million input tokens, $6 per million output tokens, $0.25 per million cached input tokens, a 1-million-token context window, and a 128,000-token maximum output, with a flat price regardless The article adds two developer-impact details not emphasized elsewhere: both Fugu Max and Fugu Ultra v2 are available only as a hosted API with no released model weights, and neither model is currently available in the EU or EEA. It also frames Sakana's separate supply-chain-resilience argument — that a swappable model

AI Weekly reports that Sakana AI shipped Fugu Max on September 11 as an orchestration engine priced at $2 per million input tokens and $6 per million output tokens, routing each request to what the company calls the leanest model capable of solving them, with pricing running 40 to 60 percent below the output cost of So On self-reported benchmarks, Fugu Max claims the best overall score on six evaluations including Terminal Bench 2.1, GPQAD, AA-LCR, GDP.pdf, AutomationBench, and SWEFish, while Fugu Ultra v2 posts 48.3 on Chartography against Opus 5's 27.3 and 74.3 on DeepSWE, without Fable 5, Fable 5.1, or GPT-6-Astra in its agent poo

Sakana AI's official release post "Introducing Fugu Max and Fugu Ultra v2: Orchestrating the Pareto Frontier" frames Fugu Max as a cost-optimized variant of the same core orchestration architecture used by Fugu Ultra v2. The post explains that Fugu Max asks "what is the best possible output we can deliver at the lowest The post summarizes the Fugu journey from an April 2026 beta through general availability, Fugu-Cyber, Sakana Chat, and the NVIDIA Nemotron integration, and asserts that a swappable pool of agents provides supply-chain resilience against vendor lock-in and revoked API access. The architectural details, family taxonomy,